Discordant response of spider communities to forests disturbed by deer herbivory and changes in prey availability Discordant response of spider communities to forests disturbed by deer herbivory and c[.]
Discordant response of spider communities to forests disturbed by deer herbivory and changes in prey availability ANDREW P LANDSMAN AND JACOB L BOWMAN1 1,2,3, Department of Entomology & Wildlife Ecology, University of Delaware, Newark, Delaware 19716 USA Department of the Interior, United States National Park Service, Hagerstown, Maryland 21740 USA Department of Biology, Hood College, Frederick, Maryland 21701 USA Citation: Landsman, A P., and J L Bowman 2017 Discordant response of spider communities to forests disturbed by deer herbivory and changes in prey availability Ecosphere 8(2):e01703 10.1002/ecs2.1703 Abstract Despite the breadth of research on impacts of dense ungulate populations and invasive plants on native vegetation, work involving indirect effects on spider communities is explicitly lacking Forest spiders depend on palatable insect prey and habitat structure, both of which are affected by herbivory and invasive vegetation To examine the indirect interactions between spiders and these influential factors, we sampled spider communities, insect prey, and vegetation in paired deer exclusion plots in central Maryland Spider abundance and richness increased with greater prey density, while increased habitat structure from deer exclusion reduced species richness and the abundance of a dominant web-building species Multivariate analyses of spider families also demonstrated the importance of both prey availability and structural complexity to spider community composition This work identifies the importance of both habitat structure and insect prey in defining the composition, abundance, and richness of forest spider communities A long history of heavy browsing pressure has resulted in local spider fauna consisting of many species that are able to thrive in low-growing vegetation and open forest understories Such changes to vegetative structure from dense deer populations and invasive plants have the potential to affect these important primary predators as well as araneophagic birds and the nutritional dynamics of forest food webs Key words: Araneae; habitat structure; indirect effects; Odocoileus virginianus; white-tailed deer Received 30 August 2016; revised 12 January 2017; accepted 13 January 2017 Corresponding Editor: Debra P C Peters Copyright: © 2017 Landsman and Bowman This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited E-mail: landsman@udel.edu INTRODUCTION insect prey if the increased structure results from greater palatable plant tissue or inflorescence density Such spatially complex habitats create a more suitable landscape upon which spiders may track down potential prey and can serve as effective refugia from araneophagic and intraguild predators (Schmidt and Rypstra 2010) Across geographic regions, increased structural complexity of vegetation concomitantly increases the diversity and abundance of spiders in all habitat types (Gibson et al 1992, McNett and Rypstra 2000) Much of the recent work involving understory spider communities and vegetative structure has The abundance of palatable insect prey and habitat structural complexity are two of the most significant predictors of spider abundance, density, and habitat selection (Rypstra 1983, Langellotto and Denno 2004) Web-building spiders are highly dependent on vegetative architecture for web attachment, while many active hunting spiders that not utilize webs for prey capture still require vegetative structure upon which to hunt for prey Structurally complex vegetation may also support increased densities of herbivorous ❖ www.esajournals.org February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN deer populations also reduces the density of native tree seedlings and other understory plants, ultimately altering entire forest plant communities (Rooney 2001, Eschtruth and Battles 2008) Certain invasive plant species also have the capacity to alter vegetative structure, leading to the potential for latent, secondary impacts on multiple trophic levels and taxa (Crooks 2002) Further obfuscating these interactions, deer browse may also facilitate the growth of some exotic species in this area of the country (Shen et al 2016) Despite near ubiquitous pressure from dense deer populations and invasive and exotic vegetation throughout forest patches in the eastern United States, knowledge of the impacts on spider communities and invertebrate trophic webs is explicitly lacking We studied the comprehensive effects of deer herbivory and invasive plant species on forest spider and insect prey communities, mediated by their effects on understory vegetation Utilizing long-term deer exclusion plots in three National Park Service (NPS) units in central Maryland, we examined forest spider abundance, species richness, community composition, and prey availability in habitats including and excluding deer herbivory Densities of whitetailed deer in our study area ranged from 46 to 66 deer/km2, a density similar to other protected, urban/suburban areas in the eastern United States (e.g., Lovallo and Tzilkowski 2003), yet well above the density recommended for successful forest regeneration (Tilghman 1989, Horsley et al 2003) Our objectives were to (1) analyze changes to plant structural complexity and species composition caused by dense deer populations and invasive plant species; (2) elucidate variations in insect prey availability; and (3) illustrate indirect effects of deer herbivory on understory spider communities mediated through forest vegetation and insect densities This provides a better understanding of the depth and complexity of these indirect interactions, particularly as this system is poorly studied despite the ubiquity of ecological pressures from deer browse and invasive plants occurred in Japanese evergreen forests, with results relatively concordant among studies: Forest understory structure increases spider richness, density, and abundance (Miyashita et al 2004) Non-predatory insects are the main source of energy and nutrients for most spiders; therefore, the availability and abundance of suitable insect prey is also an essential component of preferred habitat for spiders (Katagiri and Hijii 2015) In addition, such elevated densities of palatable insect prey also result in improved individual spider health as measured by increased biomass, longevity, and fecundity (Wise 1979, Leborgne and Pasquet 2005) As the individual success of a spider is closely related to its nutritional intake, any change in the available prey base may significantly reduce or increase the viability of a species’ population Significant changes to the abundance and nutrient dynamics of this predatory group resulting from altered habitat structure or prey base may have substantial ecological and trophic implications on any natural system, particularly as spider communities may influence the density of insect populations and provide critical nutritional contributions to araneophagic predators such as forest birds (Moulder and Reichle 1972, Gunnarsson 2007) Two common and widespread influences to both vegetative structural complexity and insect populations in the eastern United States are invasive plants and browsing from white-tailed deer (Odocoileus virginianus) As an almost exclusively insectivorous group, interactions between forest spider communities and both deer herbivory and invasive vegetation are intuitively indirect, mediated by their insect prey and the structure of vegetation in the forest understory Herbivory from dense deer populations can decrease the complexity of plant architecture by reducing stem and individual plant abundance, but may also increase individual plant structure by stimulating lateral stem and leaf development (Katagiri and Hijii 2015) Extensive deer browsing may also decrease the abundance and diversity of both predatory insects and their phytophagous insect prey (Allombert et al 2005), as well as vertebrates that rely on understory structure browsed by deer, such as forest-obligate birds and small mammals (Byman 2011, Tymkiw et al 2013) Beyond the physical structure of forest vegetation that may affect spider habitat, browsing from dense ❖ www.esajournals.org METHODS Study area We studied impacts of altered vegetative structure on spider communities in three NPS units in February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN Fig Map of study area in Washington and Frederick Counties, Maryland Green shading indicates forested land cover and black lines represent the U.S National Park Service park boundaries deer populations, paired exclusion plots were installed in 2003 by NPS and Smithsonian Institution staff throughout forests in the parks Each plot was 25 m2 with fenced plots consisting of 2.4 m tall, 10 10 cm woven wire fencing All fenced plots were open from above and permitted entry to birds and small mammals A random number generator defined the azimuth for direction of open control plots, established a distance of 10 m from fenced plots All paired plots were located at least 20 m from the forest edge In total, 25 paired plots in nine forest patches were sampled We sampled 11 paired plots in four forest sites at Antietam (Cunningham, Sherrick, Snavely, and West Woods), five paired plots in two forest sites at Monocacy (Brooks Hill and Lewis), and nine paired plots in three forest sites at C&O Canal (North Potomac, Snyders, and Taylors) central Maryland: Antietam and Monocacy National Battlefields (Antietam; Monocacy) and the C&O Canal National Historical Park (C&O Canal; Fig 1) These parks were located in a mosaicked landscape consisting of urban and suburban development, small forest fragments, and agriculture In addition to providing habitat for dense populations of white-tailed deer, forests in all parks were heavily influenced by invasive, exotic vegetation including Japanese stiltgrass (Microstegium vimineum), multiflora rose (Rosa multiflora), and garlic mustard (Alliaria petiolata) Dominant woody vegetation included native maples (Acer negundo and Acer saccharinum) and hickory species (Carya spp.), with an understory of mostly northern spicebush (Lindera benzoin) and paw-paw (Asimina triloba) To monitor changes in vegetation resulting from white-tailed ❖ www.esajournals.org February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN Fig To estimate vegetative structure in sample plots, we (A) photographed the profile board in the center of the plot, (B) grouped individual pixels into one of 15 classes based on multivariate red, blue, and green coloration, (C) reclassified pixel groups into either structure or space, and (D) estimated structure using the proportion of pixels with vegetation and the total number of pixels within the outline of the profile board Vegetation sampling height of 1.0 m For analyses, we imported images as non-georeferenced raster files in ArcGIS (ArcMap 10.0; ESRI, Redlands, California, USA) Using only the spatial extent of the board within the photograph, we then used maximumlikelihood estimation to assign each pixel to one of 15 multivariate classes based on red, blue, and green pixel coloration For a structure metric, we calculated the proportion of pixels obscured by dead and live vegetation to total number of pixels in the clipped photograph (Fig 2) We averaged the four structure estimates of vegetative structure between 0.5 and 2.0 m for each plot and used the mean value in analyses We sampled vegetation in all paired plots in July 2013 using four subplots, each m2 in size and located 0.75 m from the edges of each of the four plot sides Though vegetative response to deer exclusion was measured only during 2013, these data were included in the long-term monitoring dataset maintained by the NPS Within subplots, we identified all plants and measured the height of all seedlings, saplings, and shrubs up to 2.0 m Trees taller than 2.0 m were identified but not measured To assess the structural complexity of understory vegetation, we constructed a vegetative profile board from 0.64 cm plywood, 2.0 m in height and 35.0 cm wide We placed the board in the center of each plot and used a 12-megapixel digital camera (Nikon COOLPIX S3100; Nikon Corporation, Tokyo, Japan) to take photographs of the board from each square plot edge at a ❖ www.esajournals.org Arthropod sampling We sampled both insect and spider communities in August 2013, the time at which most web-building spiders are mature Although February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN invertebrate population densities often vary between years, we were interested in relationships with environmental parameters and not temporal differences Spiders generally develop later in the season and will similarly respond with an increase or decrease in the insect population, so we did not expect annual fluctuations to alter our effect sizes or any correlative relationships We collected arthropods using vacuum sampling methodology (Craftsman 25 cc blower/vacuum with 5-gallon paint strainer bag; Sears Brands, LLC, Hoffman Estates, Illinois, USA) for 10 within each plot We randomly sampled all vegetative surfaces for insects and hunting spiders, as well as interstitial spaces for flying insects and web-building spiders, between 0.5 and 2.0 m to encompass the browsing area for white-tailed deer Larger spiders that were not effectively collected using a vacuum sampler (e.g., adult Neoscona spp.) were collected by hand As ground spider fauna have exhibited varied responses to deer browsing, this specific group was avoided by sampling above 0.5 m After the vacuuming period, we tied off the paint strainer bag, euthanized specimens with ethyl acetate, and placed samples in 95% EtOH We identified insects to order and spiders to the lowest taxonomic ranking possible, then grouped spiders in each plot to obtain family-level dry biomass Insect prey abundance and biomass.—Insects were classified by order or functional group when great differences existed in more refined phylogenetic groups: Hymenoptera was divided into ants and parasitic wasps, Hemiptera was divided into predatory families (Reduviidae and Nabidae) and phytophagous families, and Lepidoptera was divided into larvae and adults For relevant analyses, we separately classified insects palatable to the forest understory spiders we collected (Diptera, non-predatory Hemiptera, Psocodea, and Thysanoptera) (Nentwig 1983, Bardwell and Averill 1997) We used mixed-effect models in package “nlme” for analyzing relationships between insect abundance and biomass, native and exotic plant richness, and the exclusion of deer herbivory (Pinheiro et al 2016) We used a logarithmic transformation to obtain normality for insect groups that were non-normally distributed Spider communities.—We classified spiders into five functional groups based on prey capture method, namely orb web-weaving spiders, space or cobweb weavers, active hunting species, ground-dwelling spiders, and species that almost exclusively utilize other spiders for prey Collected only incidentally in understory vegetation, ground spiders were identified but excluded from analysis and the kleptoparasitic and araneophagic species were excluded from multivariate analyses as they represented only a small portion of overall spiders collected As mixed-effect models with spider counts exhibited non-constant error variance, we used generalized linear models (GLM) in packages “lme4” and “MASS” to analyze the response of spider abundance and richness to deer exclusion, insect prey abundance, and forest patch (Venables and Ripley 2002, Bates et al 2015) We also incorporated our metric of vegetative structure into subsequent models, but did not include deer exclusion and vegetative structure in the same models due to their covariance Depending on model overdispersion, as tested using function “dispersiontest” in R package “AER,” we used either a Poisson or negative binomial probability distribution (Kleiber and Zeileis 2008) Species diversity indices for spider communities were not used in analysis as too few adults were captured to effectively represent the community at the species or genus level To comprehensively assess changes to spider community structure, we used multivariate linear regression to model Statistical analyses Vegetative diversity and structure.—We calculated Shannon–Wiener diversity of woody plant species and used the Jaccard dissimilarity metric to compare both woody and herbaceous species between plots and forest patches We then used linear mixed-effect models to assess the impacts of deer exclusion on vegetative diversity, species richness, and structure To account for broad differences in vegetative communities between forest patches, individual patch was included as a random effect in mixed models We utilized the Jaccard index to perform permutational multivariate analysis of variance (MANOVA) of plant community dissimilarity between forest patches to ensure the appropriate strata for the random effect (McArdle and Anderson 2001) Though these permutational models are less sensitive to heterogeneity, we used “betadisper” in package “vegan” to ensure multivariate homogeneity of variances (Oksanen et al 2016) ❖ www.esajournals.org February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN effects of deer exclusion, insect prey, and forest patch on the relative abundance of spider functional groups and families We also performed post hoc univariate linear regression on the relative abundance of functional groups and used Poisson GLM for analyzing the raw abundance of individual families Bray-Curtis ecological dissimilarity of spider communities was used in permutational MANOVA tests for both functional groups and taxonomic families to assess variations in spider community response to deer exclusion, prey abundance, and forest patch Analogous to the permutational MANOVA, we used distance-based as well as traditional redundancy analysis (RDA) to plot ordination figures for spider communities For all multivariate tests, we used a factorial variable derived from our measure of insect density by splitting the variable distribution into seven classes in increments of 50 individuals All statistical analyses were completed in Program R 3.3.1 (R Core Team 2016) Fig Box plot showing effect of deer exclusion by fencing on vegetative structure in the three parks Boxes represent first and third quartiles and median Points greater than 1.5 times the interquartile range are depicted as outliers RESULTS Vegetative diversity and structure 225.96 13.48 insects per plot, with 145.75 10.48 palatable insects Insect orders exhibited various responses to changes in nativity and structure of vegetation, depending on order and typical diet (Table 1) Overall insect abundance decreased with exotic plant species richness (t = 2.327, P = 0.025) Palatable insect prey abundance and biomass also decreased with exotic plant species richness (P = 0.031; P = 0.036) Though deer exclusion was not a significant predictor of either insect prey or overall invertebrate abundance, removal of deer herbivory did increase both the abundance and biomass of insect herbivores (P = 0.040; P = 0.045) Vegetative structure increased insect prey abundance only at a = 0.1 (t = 1.693, P = 0.099) Native plant species richness increased the abundance or biomass of several insect orders, including flies (P = 0.097), beetles (P = 0.074), and caterpillars (P = 0.060) Conversely, greater exotic plant species richness decreased the abundance of flies (P = 0.009), beetles (P = 0.043), and overall insect prey (P = 0.031), as well as true bug biomass (P = 0.038) Beetle abundance and biomass increased with deer exclusion (P < 0.001; P < 0.001); however, the abundance of parasitic Throughout all plots, we identified 26 tree and shrub species, as well as two graminoid, four vine, and 34 herbaceous species Vegetative communities were dissimilar between forest patches (F8,41 = 2.707, P = 0.001) Richness of native species and all plant species did not differ in fenced and control plots (P > 0.1), with a mean of 8.26 0.41 total species and 5.66 0.38 native species per plot However, diversity of woody plant species increased with deer exclusion (t = 2.685, P = 0.011) as fenced plots contained greater evenness Plant structure was positively correlated with deer exclusion (t = 4.021, P < 0.001), though trends were not significant in plots at C&O Canal (Fig 3) Habitat structure increased with the presence of multiflora rose (t = 2.458, P = 0.019) Although habitat structure was visibly reduced in forests with Japanese stiltgrass, too few paired plots contained this species to include in models Insect prey abundance and biomass We collected 11,298 insects and related taxa, including three non-spider arachnid groups Insects from 15 orders were found, with flies, true bugs, and parasitic Hymenoptera the most abundant taxa We collected an average of ❖ www.esajournals.org February 2017 ❖ Volume 8(2) ❖ Article e01703 LANDSMAN AND BOWMAN Table Insect community response to deer exclusion and species richness of native and exotic plants Native plant richness Deer exclusion Exotic plant richness Taxon Variable t P t P t P All insects Spider prey Spider prey Herbivores Herbivores Coleoptera Coleoptera Diptera Hemiptera Hemiptera Hymenoptera (ants) Lepidoptera Lepidoptera (caterpillars) Psocodea Abundance Abundance Biomass Abundance Biomass Abundance Biomass Abundance Abundance Biomass Abundance Abundance Abundance Abundance 0.117 0.046 0.087 2.125 2.074 3.928 4.923 0.478 1.430 0.643 1.666 0.743 1.747 0.491 0.9076 0.9633 0.9316 0.0402 0.0449 0.0003